Hypomethylation and Hypermethylation of DNA in Wilms Tumors

Hypomethylation and hypermethylation of DNA in Wilms tumors

Melanie Ehrlich*,1, Guanchao Jiang1, Emerich Fiala2, Jeﬀrey S Dome3, Mimi C Yu4, Tiﬀany I Long5,6, Byungwoo Youn2, Ock-Soon Sohn2, Martin Widschwendter5,6, Gail E Tomlinson7, Murali Chintagumpala8, Martin Champagne9, David Parham10, Gangning Liang6, Karim Malik11 and Peter W Laird5,6

1Tulane Cancer Center and Human Genetics Program, Tulane Medical School, New Orleans, Louisiana, LA 70112, USA; 2Department of Biochemical Pharmacology, American Health Foundation, Valhalla, New York, NY 10595, USA; 3Department of Hematology/Oncology, St. Jude Children’s Research Hospital, Memphis, Tennessee, TN 38105, USA; 4Department of Preventive Medicine, University of Southern California, Norris Comprehensive Cancer Center, Los Angeles, California, CA 90033, USA; 5Department of Surgery, University of Southern California, Norris Comprehensive Cancer Center, Los Angeles, California, CA 90033, USA; 6Department of Biochemistry and Molecular Biology, University of Southern California, Norris Comprehensive Cancer Center, Los Angeles, California, CA 90033, USA; 7Department of Pediatrics, University of Texas Southwestern Medical Center, Dallas, Texas, TX 75235, USA; 8Texas Children’s Cancer Center, Baylor College of Medicine, Houston, Texas, TX 77030, USA; 9Service D’hematologie Oncologie, Hopital Sainte-Justine, Montreal, Quebec, H3T 1C5, Canada; 10Department of Pathology, Arkansas Children’s Hospital and University of Arkansas for Medical Sciences, Little Rock, Arkansas, AR 72205, USA; 11Cancer & Leukaemia in Childhood Unit, Department of Pathology, School of Medical Sciences, University of Bristol, Bristol, BS8 1TD, UK

We quantitatively analysed hypermethylation at CpG Keywords: DNA hypomethylation; DNA hypermethy- islands in the 5’ ends of 12 genes and one non-CpG lation; Wilms tumors island 5’ region (MTHFR) in 31 Wilms tumors. We also determined their global genomic 5-methylcytosine content. Compared with various normal postnatal tissues, *40 – 90% of these pediatric kidney cancers were Introduction hypermethylated in four of the genes, MCJ, RASSF1A, TNFRSF12 and CALCA as determined by a quantita- Numerous recent studies have described cancer-asso- tive bisulfite-based assay (MethyLight). Interestingly, the ciated hypermethylation in CpG-rich 5’ gene regions non-CpG island 5’ region of MTHFR was less (Baylin and Herman, 2000; Issa, 2000). However, net methylated in most tumors relative to the normal tissues. decreases in DNA methylation in cancer often exceed By chromatographic analysis of DNA digested to the localized increases in DNA methylation associated deoxynucleosides, about 60% of the Wilms tumors were with carcinogenesis (Ehrlich, 2000; Gama-Sosa et al., found to be deficient in their overall levels of DNA 1983). The importance of cancer-linked hypermethyla- methylation. We also analysed expression of the three tion of transcription control regions is clear because of known functional DNA methyltransferase genes. No the consequent transcriptional silencing, especially for relationship was observed between global genomic 5- tumor suppressor genes (Baylin and Herman, 2000; methylcytosine levels and relative amounts of RNA for Issa, 2000), but the biological significance of DNA DNA methyltransferases DNMT1, DNMT3A, and hypomethylation in cancers is less certain. DNMT3B. Importantly, no association was seen between Global DNA hypomethylation may be as frequent in CpG island hypermethylation and global DNA hypo- cancer as CpG island hypermethylation. It has been methylation in these cancers. Therefore, the overall found in various cancers, especially metastases, in a genomic hypomethylation frequently observed in cancers study in which cancers were compared to a wide range is probably not just a response or a prelude to of normal tissues (Gama-Sosa et al., 1983). Global ONCOGENOMICS hypermethylation elsewhere in the genome. This suggests DNA hypomethylation has also been seen in compar- that the DNA hypomethylation contributes independently isons of diverse types of cancers and adjacent, to oncogenesis or tumor progression. apparently normal tissues (reviewed in Ehrlich, 2002). Oncogene (2002) 21, 6694 – 6702. doi:10.1038/sj.onc. For ovarian epithelial tumors, a low overall genomic 1205890 DNA methylation level (relative to that of various normal tissues) has been shown to be significantly associated with the degree of malignancy (benign cystadenomas, low malignant potential tumors, or adenocarcinomas; Cheng et al., 1997; Qu et al., 1999a). *Correspondence: M Ehrlich, Human Genetics Program SL31, In this study on epigenetic changes in Wilms tumors, Tulane Medical School, 1430 Tulane Avenue, New Orleans, Louisiana, LA 70112, USA; E-mail: [email protected]. we present the first comparison of cancer-linked global Received 29 April 2002; revised 12 July 2002; accepted 18 July genomic hypomethylation and hypermethylation at 2002 multiple CpG islands in the same tumors. We found Wilms tumor DNA hypomethylation and hypermethylation M Ehrlich et al 6695 that CpG island hypermethylation was neither posi- Table 1 Global DNA methylation levels in normal tissues and tively nor negatively associated with global DNA Wilms tumors hypomethylation. Also, we showed that global DNA % C methylated a methylation was not associated with the relative levels Sample +s.d. (n) of RNA from the DNA methyltransferase genes Normal tissues DNMT1, DNMT3A and DNMT3B. Cerebellum B 4.04+0.16 (9) Cerebellum C 3.98+0.02 (3) Heart B 3.43+0.13 (9) Results Heart C 3.44+0.04 (3) Kidney B 3.77+0.06 (9) Kidney A 3.73+0.06 (9) Frequent global DNA hypomethylation in Wilms tumors Lung B 3.66+0.12 (12) As in our previous studies of differences in human Lung C 3.68+0.06 (3) Spleen B 3.80+0.05 (6) DNA methylation levels among normal human tissues Spleen C 3.77+0.05 (6) and between normal tissues and cancers (not including Kidney adj WT2 3.93+0.16 (6) Wilms tumors), the m5C content was determined on Kidney adj WT29 3.90+0.07 (6) DNA digested to deoxynucleosides (Ehrlich et al., 1982; Gama-Sosa et al., 1983). The choice of normal WT DNAs with strong global hypomethylation 5 WT4 2.88+0.15 (3) tissue for m C content comparisons is complicated by WT8 3.15+0.07 (6) the fact that Wilms tumors are derived from the WT9 3.09+0.14 (5) embryonic metanephric blastema, which is usually WT10 3.10+0.07 (5) eliminated during embryonic development and, if not, WT14A 3.18+0.04 (2) WT20 3.17+0.07 (8) in infancy. Furthermore, histologically normal kidney WT30 3.08+0.09 (3) tissue of Wilms tumor patients may contain foci of WT31 2.94+0.05 (3) abnormal cells. Because normal kidney is not an WT33 3.18+0.13 (3) appropriate control for Wilms tumors, we used a panel of adult normal human tissues for comparison. We set WT DNAs with moderate global hypomethylation WT2 3.35+0.01 (2) the criterion for global hypomethylation of Wilms WT3 3.40+0.06 (2) 5 tumor DNA as lower levels of genomic m C than seen WT6 3.20+0.01 (2) in the normal postnatal somatic tissue with the lowest WT7 3.28+0.02 (3) m5C content in its DNA (heart DNA, Table 1). Out of WT15 3.34+0.13 (5) WT18 3.38+0.10 (5) 31 tumors obtained before treatment, 19 displayed WT22 3.39+0.07 (3) global DNA hypomethylation by this criterion. There WT24 3.28+0.06 (3) was no significant association of global DNA hypo- WT29 3.25+0.08 (3) methylation with tumor stage. WT32 3.35+0.02 (3) WT DNAs with no global hypomethylation CpG island hypermethylation in Wilms tumors WT1 3.57+0.01 (2) WT5 3.48+0.06 (6) We used MethyLight, a bisulfite modification-depen- WT11 3.62+0.04 (2) dent, fluorescence-based real-time PCR assay (Eads et WT13 3.55+0.11 (5) WT16 3.67+0.09 (5) al., 2000), to analyse CpG island methylation in Wilms WT17 3.59+0.05 (5) tumors. Controls of unmethylated DNA and in vitro WT21 3.90+0.03 (3) methylated DNA were included in each analysis. WT23 3.46+0.14 (3) Methylation is detected when template sequences WT26 3.55+0.10 (3) corresponding to both of the primers and the probe WT27 3.46+0.06 (3) WT28 3.44+0.06 (3) for a given CpG island are methylated at multiple WT34 3.52+0.12 (3) CpGs. From 5 (e.g. ICAM1) to 11 (e.g. CXADR) CpGs were analysed per reaction. To adjust for aThe mean number and standard deviation for cytosine methylation differences in genomic DNA integrity and quantity, levels are given and, in parentheses, the number of HPLC determinations for each sample is shown. Strong and moderate each methylation measurement was normalized against hypomethylation, 53.20 and 3.20 – 3.40% of C residues methylated, PCR results from the same sample at a DNA sequence respectively. No global hypomethylation, % C methylation within the containing no CpGs. This ratio was then further range of the normal tissues normalized to a common reference sample, in vitro methylated sperm DNA. The resulting value is expressed as a percentage of this methylated reference to the quantitative combined bisulfite restriction (PMR, see Materials and methods). PMR values can analysis and bisulfite-based genomic sequencing (Eads exceed 100, if the reference DNA sample was not fully et al., 2000). methylated in vitro. This does not affect comparisons We initially screened DNA from 10 Wilms tumor among various test samples, which are all made relative samples for hypermethylation at 30 different CpG-rich to the same reference sample. This method has been gene regions. Seventeen of these loci showed little or no extensively validated as to its specificity, sensitivity, methylation and were not considered for further reproducibility, and quantitative nature by comparison analysis. These included CDKN2A, APC, ARF,

Oncogene Wilms tumor DNA hypomethylation and hypermethylation M Ehrlich et al 6696 DAPK1, GSTP1, MLH1, TGFBR2, TIMP3,and these genes, respectively. The tumor samples in Tables MYOD1 (data not shown). We analysed the methyla- 1 and 3 were from previously untreated patients and tion status of the remaining twelve CpG islands and included a tumor obtained before chemotherapy from the 5’ region of the MTHFR gene in 12 normal tissue one kidney of a patient (WT10). We also analysed a samples and 31 Wilms tumors. The genes tested tumor obtained from the other kidney of this patient included the following tumor suppressor, drug resis- after chemotherapy (WT25). These bilateral tumors tance, or candidate tumor suppressor genes: differed in CpG island hypermethylation. WT10 TNFRSF12 (tumor necrosis factor receptor super- displayed hypermethylation in both RASSF1A and family, member 12; also known as TRAMP or DR3), MCJ, but WT25 exhibited hypermethylation in ESR1 (estrogen receptor a), MADH3 (SMAD3), RASSF1A and TNFRSF12 instead of MCJ (Table 3 MGMT1 (O6-methylguanine-DNA methyltransferase), and data not shown). Also, WT10 had much global ABCB1 (MDR1; multidrug resistance gene 1), MCJ DNA hypomethylation while WT25 had DNA methy- (methylation-controlled J protein), CCND2 (cyclin lation levels within the normal range (Table 1 and data D2), and RASSF1A (Ras association domain family not shown). Similarly, WT10 had much more satellite 1A; Tables 2 – 4). We also analysed the 5’ CpG islands DNA hypomethylation than WT25 (Ehrlich et al., of CALCA, a cancer hypermethylation marker gene 2002). (Issa, 2000); PTGS2 (cyclooxygenase 2), a candidate We had sufficient amounts of sample to analyse proto-oncogene (Toyota et al., 2000); ICAM1 (intra- methylation of the RASSF1A, TNFRSF12, and MCJ 5’ cellular adhesion molecule 1), a gene that is CpG islands in two Wilms tumor patients for whom upregulated in nasopharyngeal carcinoma (Busson et matched sets of nephrogenic rest, tumor, and normal al., 1992); and CXADR (Homo sapiens coxsackie virus post-natal kidney samples were available. Nephrogenic and adenovirus receptor). The only 5’ region that did rests are precursor lesions associated with Wilms not satisfy the formal criteria for CpG islands tumors and, like Wilms tumors, contain blastemal cells (Gardiner-Garden and Frommer, 1987) was that of (undifferentiated mesenchymal cells). They are often MTHFR (methylene tetrahydrofolate reductase) (Eads found in kidneys of Wilms tumor patients. Both et al., 2001). Interestingly, this 5’ region was the most nephrogenic rest samples showed negligible methyla- consistently methylated at high levels among the tion of the RASSF1A and MCJ CpG islands unlike the normal tissues analysed (Table 2). Most of the 12 corresponding Wilms tumors, which were extensively CpG islands exhibited little or no CpG methylation in hypermethylated in these regions (data not shown). the normal tissues examined (Table 2). However, Furthermore, in contrast to Wilms tumors, we MGMT and TNFRSF12 displayed appreciable methy- observed negligible levels of RASSF1A, TNFRSF12 lation in many of the normal tissue samples and and MCJ CpG island methylation in a 22-week fetal RASSF1A showed tissue-specific differences in methy- kidney DNA. Therefore, hypermethylation of RASS- lation. F1A, TNFRSF12 and MCJ CpG islands is tumor- Significant hypermethylation of RASSF1A, specific and not merely reflective of an oncofetal TNFRSF12, MCJ and CALCA CpG islands expansion of blastemal cells, embryonic kidney precur- (P50.0005, P50.01, P50.0001 and P50.09, respec- sor cells. tively) was seen in Wilms tumors (Table 3, bold There are CpG-rich regions that are appreciably typeface PMR values) relative to various normal methylated in a variety of normal somatic tissues tissues, including morphologically normal tumor-adja- (Zhang et al., 1987; Strichman-Almashanu et al., 2002). cent tissue (Table 2). About 75, 65, 90 and 40% of the The MethyLight assay revealed statistically significant tumors were hypermethylated in the 5’ CpG islands of hypomethylation of one such region, the 5’ end of

Table 2 The extent of methylation of 13 5’ CpG-rich sequences in normal tissues Extent of methylation of the indicated CpG-rich sequence (PMRa) Sample, age, sex ABCB1 CALCA CCND2 CXADR ESR1 ICAM1 MADH3 MGMT PTGS2 RASSF1A TNFRSF12 MCJ MTHFR

Cerebellum B, 19 years, M 8 0 0 0 0 5 2 10 0 2 26 1 12 Cerebellum C, 68 years, M 4 0 0 1 0 7 3 3 0 3 10 0 16 Heart B, 19 years, M 5 0 0 0 0 9 2 11 0 0 25 1 2 Heart C, 68 years, M 3 0 0 0 1 7 1 8 0 NA 5 NA 11 Kidney B, 19 years, M 7 0 0 0 0 13 1 32 0 20 24 4 16 Kidney A, 56 years, M 7 0 0 0 0 6 1 9 1 24 24 1 21 Lung B, 19 years, M 8 0 0 0 0 8 5 9 0 0 19 3 42 Lung C, 68 years, M 8 0 0 0 0 9 2 6 1 0 5 1 24 Spleen B, 19 years, M 7 4 0 0 1 8 3 5 3 0 6 4 72 Spleen C, 68, M 2 5 0 0 0 2 1 2 3 0 4 1 46 Kidney adj WT24, 5 years, M 2 0 0 3 0 2 0 17 0 14 12 2 25 Kidney adj WT29, 2 years, M 3 NA 0 0 0 4 0 14 0 5 2 4 10

aThe PMR values, which indicate the extent of methylation of the sample relative to a reference, are given for autopsy tissues from trauma victims, except for the last two samples, which were from apparently normal kidney adjacent to the indicated Wilms tumor sample. NA, not assayed

Oncogene Wilms tumor DNA hypomethylation and hypermethylation M Ehrlich et al 6697 Table 3 The extent of methylation of 13 5’ CpG-rich sequences in Wilms tumors WT sample, age, sex, Extent of methylation of the indicated CpG-rich sequence (PMRa) tumor stage ABCB1 CALCA CCND2 CXADR ESR1 ICAM1 MADH3 MGMT PTGS2 RASSF1A TNFRSF12 MCJ MTHFR

WT DNAs with strong global hypomethylation WT4 3 years F, III 14 0 0 0 0 45 9 4 0 5 50 2 7 WT8 1 year F, II 9 0 1 0 1 16 1 7 0 151 29 73 5 WT9 1 year F, II 7 50 00012NA085 35 133 15 WT10 9 years M, V 7 5 0 0 0 13 1 32 0 58 7 53 6 WT14A 5 years F, III 3 0 0 0 0 1 3 4 0 36 52 38 3 WT20 1 year F, III 7 51 220146082 47 149 12 WT30 4 years F, II 8 5 0 0 0 12 5 5 0 456 30 184 8 WT31 2 years M, III 5 NA 0 0 0 5 NA 4 0 102 0 11 9 WT133 1 year F, III 0 0 0 0 0 1 0 7 0 1 0 101 1

WT DNAs with moderate global hypomethylation WT2 3 years F, III 8 0 0 0 0 24 0 1 0 3 44 140 2 WT3 3 years M, III 19 0 0 0 0 13 3 3 1 117 48 102 9 WT6 9 months F, II 9 0 0 NA 63 2 8 6 0 74 636 WT7 8 months M, IV 4 0 0 0 0 13 1 9 0 18 3 23 3 WT15 2 years M, I 2 110 00030110162 29 129 5 WT18 1 year M, II 5 NA 0 0 0 1 2 13 0 115 43 114 7 WT22 4 years M, I 10 NA 0 0 0 16 5 9 0 248 71 113 WT24 5 years M, I 1 0 0 0 0 2 2 9 0 167 31 63 2 WT29 2 years M 4 127 0000NA9179 40 221 1 WT32 11 months F, II 1 71 0007070117 30 64 14

WT DNAs with no global hypomethylation WT1 5 years F, II 5 5 NA 0 0 3 10 10 0 NA 43 29 8 WT5 3 years F, I 10 NA 0 0 0 25 1 1 0 85 73 225 2 WT11 7 years M, III 1 11 0 0 0 5 1 26 0 212 11 123 1 WT13 10 months M, I 4 1 0 0 0 8 0 4 0 1 16 55 2 WT16 4 years F, II 2 88 001518085 54 92 2 WT17 10 months M, I 2 0 0 0 1 3 1 38 0 0 18 11 5 WT21 4 years F, II 15 56 5002000264 98 29 0 WT23 2 years F, V 7 0 0 0 0 1 5 5 0 19 49 163 1 WT26 2 years M, I 3 5 0 0 0 11 1 22 0 7 17 39 14 WT27 5 years M, III 5 0 3 0 0 0 0 2 0 94 79 170 9 WT28 3 years M, III 5 0 0 0 1 11 2 12 0 53 47 97 8 WT34 5 years F, III 1 11 00051260212 11 123 1 aThe extent of methylation is given in PMR values as for Table 2. For genes with signiﬁcant cancer-associated hypermethylation, PMR values indicating hypermethylation relative to control tissues (Table 2) are shown in boldface. For MTHFR, the samples that were hypomethylated are in bold and italics. NA, not assayed

MTHFR, in Wilms tumors compared to normal tissues 1p36.2-p36.3, and 11p15.2-p15.1, respectively. The only (P50.001; Tables 2 and 3). In contrast to the other test genes for which this correction was necessary were examined 5’ CpG-rich regions, which showed either MCJ and RASSF1A. With respect to RASSF1A on increased methylation or no change in tumors 3p21, two of the tumors (WT5 and WT6) with an extra compared to control tissues, MTHFR’s CpG-rich copy of chromosome 3 showed high PMR values for region does not fulfil the definition of a CpG island, the RASSF1A CpG island indicative of extensive as mentioned above. It is not as highly enriched in methylation in this region. Similarly, the seven tumors CpG’s (observed/expected CpG, 0.56; 56.5% G+C) that exhibited gains of chromosome 13 (#s 1, 11, 14A, and encompasses only about 200 bp. 15, 16, 27, and 33) showed higher PMRs for the MCJ The PMR values for the extent of methylation are CpG island methylation at 13q14 (Table 3) than the obtained from a ratio of measurements of the gene of normal somatic tissues (Table 2). For MethyLight, the interest and of a CpG-free DNA locus elsewhere in the CpG-free standard sequence was in ACTB at 7p15-p12. genome. Cytogenetic abnormalities involving these From the cytogenetic data, its copy number was regions would affect this ratio, independently of deduced to be abnormal for four tumors, namely, methylation status. We had analysed the tumor WT15 and WT30 (copy number 3), WT7 (copy number samples cytogenetically (Ehrlich et al., 2002). There- 1), and WT1 (average copy number 3.2). The PMR fore, we checked whether the PMR values for the five values reported in Table 3 have been corrected for all genes with the most significant tumor-linked methyla- of these alterations in copy number. tion differences (MCJ, RASSF1A, MTHFR, We also compared DNA hypermethylation and TNFRSF12 and CALCA) required adjustment due to hypomethylation. We found no significant associations chromosome arm imbalances in any of the tumors. between the methylation status of any of the 13 studied These genes are located at 13q14.1, 3p21.3, 1p36.1, CpG-rich regions and global methylation levels (Table

Oncogene Wilms tumor DNA hypomethylation and hypermethylation M Ehrlich et al 6698 1) or satellite DNA hypomethylation (satellite a or

., 2001) chromosome 1 satellite 2; Ehrlich et al., 2002). In addition, there was no signiﬁcant association between et al tumor stage and CpG island methylation status.

Comparison of DNMT RNA levels in different Wilms tumors We investigated whether global DNA hypomethylation was associated with reduced expression levels of DNMT1, DNMT3A,orDNMT3B in 15 Wilms tumors, as measured by real-time RT – PCR (Eads et al., 1999). Since we do not have an appropriate control tissue against which to compare DNMT expression levels in Wilms tumors, we compared relative expression levels among the diﬀerent Wilms tumor samples, to see whether the tumors with the most severe global hypomethylation had the lowest levels of DNMT gene expression. The expression levels for DNMT1, DNMT3A, and DNMT3B were normalized against

a those of two other proliferation genes, PCNA or H4F2, with similar results (Figure 1 and data not shown). The examined tumor DNAs included samples that had very low global DNA methylation levels (3.08 – 3.20% C residues methylated) as well as those in the range of normal tissues (3.40 – 3.90% C residues methylated). We did not observe an association between global DNA hypomethylation and the amounts of RNA for any of the DNMT genes (P=0.51, 0.32, and 0.44 for genomic m5C levels and DNMT1, DNMT3A, and DNMT3B expression, respectively). Furthermore, we did not observe a signiﬁcant association between Parameters for new genes analysed by MethyLight , and primer and probe sequences for MethyLight analysis of these regions were previously described (Eads PTGS2 Table 4 , and UTR 0.79 TTTCGGGTCGTTTTGTTATGG ACTACAAATACTCAACGTAACGCAAACT 6FAM-TCGCCAACTAAAACGATAACACCACGAACA-BHQ-1 ’ MTHFR , ) exon) CpG:GpC Forward primer sequence Reverse primer sequence Probe oligo sequence 34 Promoter 0.52 TCGGGTCGGGAGTAGTTATTTG CGACTATACTCAACCCACGCC 6FAM-ACGCTATTCCTACCCAACCAATCAACCTCA-BHQ-1 87 Promoter 0.78 GGTTAGCGAGGGAGGATGATT TCCCCTCCGAAACAAATACTACAA 6FAM-TTCCGAACTAACAAAATACCCGAACCGAAA-BHQ-1 1126 Promoter 0.89 GGAGGGTCGGCGAGGAT TCCTTTCCCCGAAAACATAAAA 6FAM-CACGCTCGATCCTTCGCCCG-BHQ-1 150 Promoter 1.00 TACGCGGTTGGAGAAGTCG ATAAACTCGCGTCACTTCCGA 6-FAM-AACGACCCGAACCGAACTACGAACG-BHQ-1 bp 7 7 7 7 MGMT 112/ 166/ location Location of 216/ Amplicon , 1189/ 7 7 7 7 ) start ( bp ESR1

, Figure 1 Relative expression levels of DNMT1, DNMT3A and DNMT3B in Wilms tumors. Steady-state mRNA levels of DNMT genes were determined from total RNA preparations of 15 Wilms

CALCA tumors (WT1, 2, 5, 6, 9, 11, 14A, 20, 21, 23, 27, 28, 29, 30, 33) using real-time ﬂuorescence-based RT – PCR (TaqMan). The expression levels were calculated as the ratio between the DNMT gene measurement and each of the reference gene measurements (average of H4F2 and PCNA), to correct for variations in the amounts of RNA. H4F2 and PCNA are cell-cycle regulated like

GenBankaccession Amplicon location transcription (e.g., promoter, relative to amplicon in gene the DNMT genes. The reference gene-normalized DNMT expression levels of Wilms tumors are shown relative to the average expression level for all tumors. For every examined tumor, these relative values for each DNMT measurement were plotted against The CpG-rich regions of ABCB1 L07624 929 – 1007 a GeneCCND2 number (GenBankCXADR numbering, U47284 AF242862 3800 – 3866 281 – 344 TNFRSF12 AB051850ICAM1 M65001 27 – 95 1187 – 1266 +27/+95 Exon 0.70 GCGGAATTACGACGGGTAGA ACTCCATAACCCTCCGACGA 6FAM-CGCCCAAAAACTTCCCGACTCCGTA-BHQ1 MADH3MCJ U68019RASSF1A AC002481 AF126743 18107 – 99 18171 – 173 407 – 487 +57/+121 Promoter/Exon +99/+173 1 +407/+487 0.71 Exon ATTGAGTTGCGGGAGTTGGT 5 the global 0.86 ACACGCTCCAACCGAATACG methylation CGTGAAGCGTTTGTTGGGT TTAACCGCCTTCTCGCACC 6FAM-CCCTTCCCAACGCGCCCA-BHQ-1 levels 6FAM-TCCTCCTACCCGTTCTACTCGCCCTTCTT-BHQ-1 (% C methylated)

Oncogene Wilms tumor DNA hypomethylation and hypermethylation M Ehrlich et al 6699 methylation levels at any of the 13 gene regions and the genome than in the least methylated normal expression of any of the three DNMT genes (39 postnatal somatic tissue in the controls. Therefore, we individual comparisons). may underestimate hypomethylation in tumors that have only a small degree of DNA hypomethylation. Nonetheless, about 60% of the tumors were found to Discussion be hypomethylated. Previously, Wilms tumor-specific gene-region hypo- We compared DNA hypomethylation and hypermethy- methylation was reported only for several imprinted lation in Wilms tumors, a type of cancer whose regions (Malik et al., 2000; Tycko, 2000). Comparison composition of 580% cancer cells is well suited for in the MethyLight assay of various normal tissue this type of study. Frequent hypermethylation of some DNAs with Wilms tumor DNA samples revealed a CpG islands was observed (Table 3), as is the case for region of significant tumor-specific hypomethylation, most kinds of cancer. Previously, regional cancer- namely, the CpG-rich 5’ end of MTHFR (P50.001; linked alterations in methylation in Wilms tumors have Table 3). MTHFR is on a subband of lp36 neighboring been reported for gene regions associated with that of TNFRSF12. It is widely expressed, but its RNA imprinting, the WT1 gene, and the CpG island in the is present at only low levels (Gaughan et al., 2000). As 5’ region of CDKN2A (p16) (Arcellana-Panlilio et al., described above, its partially CpG-rich 5’ region does 2000; Malik et al., 2000; Mares et al., 2001; Tycko, not fulfil the typical criteria of a CpG island and this 2000) and recently for RASSF1A (Harada et al., 2002). may be why it escaped the hypermethylation seen for In the present study, comparing Wilms tumors and some of the other examined 5’ gene regions. normal tissues, there was highly significant hyper- For several reasons, it does not appear that the methylation of CpG islands near the 5’ ends of MCJ, observed tumor-associated alterations in DNA methy- RASSF1A, and TNFRSF12 as well as significant lation can be ascribed to the young age of most of the hypermethylation of a tumor marker gene CALCA patients and age-related changes in DNA methylation. (Tables 2 and 3). Of particular interest is the MCJ No consistent differences were apparent in the CpG island, which was hypermethylated in about 90% methylation status of any of the studied DNA of the Wilms tumors, with very low levels of CpG sequences between control tissues from an individual island methylation in normal tissues. This gene encodes less than 20 years and one more than 60 years (Table a protein related to the E. coli heat shock protein 2). These are age cut-offs for which age-related DNA DNAJ, and may play a role in enhanced sensitivity to hypermethylation differences were previously estab- cancer chemotherapeutic drugs. The only previous lished (Issa, 2000). Age-dependent changes in DNA report of aberrant methylation of this newly identified methylation have been observed between young and gene was about hypermethylation of its 5’ CpG island old individuals but not between individuals much in ovarian carcinomas, in which LOH of one allele and younger than 20 vs those aged 20 – 30 years (Ahuja et reduced or lost expression of MCJ was often seen al., 1998; Issa et al., 1996). These changes usually (Shridhar et al., 2001). The second most frequently pertain to CpG island hypermethylation (Issa, 2000), hypermethylated CpG island observed in the Wilms although not always (Yung et al., 2001). One of the tumors was in the 5’ region of RASSF1A (Tables 3 and CpG islands for which increasing age was found to be 4). This gene is implicated as a suppressor of kidney associated with more methylation is that of ESR1 tumor formation (Dreijerink et al., 2001) and is (Ahuja et al., 1998). Ahuja et al. (1998) demonstrated frequently hypermethylated in clear cell renal carcino- an age-dependency for increased methylation of this 5’ mas as well as various other cancers (Burbee et al., island both in normal liver and colon but, nonetheless, 2001). In this study we compensated for increased copy they observed significant tissue-specific differences. In numbers of genes due to numerical aberrations or our study of five types of normal tissue that had not partial chromosome duplications in the tumors. Gains been analysed for age-dependency of ESR1 methyla- in chromosomes containing MCJ or RASSF1A were tion by Ahuja et al. (1998), we found no detectable usually accompanied by hypermethylation of the 5’ changes in methylation with age (Table 2). To further CpG islands of these genes. This suggests that address the question of whether age-related differences promoter region hypermethylation sometimes mitigates in DNA methylation contributed to our results, we the cancer-antagonistic effects of extra copies of tumor- compared Wilms tumor kidney tissue from two patients suppressor genes in aneuploid tumors. with adjacent kidney samples that contained no signs We also report for the first time that most Wilms of nephroblastomatosis. In the comparison of these tumors display global genomic hypomethylation (Table age-matched tissues, we again observed cancer-asso- 1), as has been described for some other types of ciated CpG island hypermethylation (Tables 2 and 3). cancers (Feinberg and Vogelstein, 1983; Gama-Sosa et Furthermore, these alterations in methylation in al., 1983; Ehrlich, 2000). Given the tissue-specific Wilms tumors are not just due to the embryonic differences in overall m5C levels in human DNA nature of the tissue from which these pediatric cancers (Ehrlich et al., 1982) (Table 1) and the origin of Wilms arise. First, in contrast to Wilms tumors, we observed tumors from embryonic tissue, we compared tumors to negligible levels of RASSF1A and MCJ CpG island a wide variety of postnatal somatic tissues. We define methylation and no satellite DNA hypomethylation tumor DNA hypomethylation as less methylation in (data not shown) in DNA from nephrogenic rests

Oncogene Wilms tumor DNA hypomethylation and hypermethylation M Ehrlich et al 6700 (Wilms tumor precursors) or 22-week fetal kidney. the association between the two has not been Furthermore, the global DNA demethylation, includ- previously investigated for any tumor type. Our study ing satellite DNA demethylation that occurs from the indicates that these two cancer-associated mechanisms one-cell through the morula stage is reversed by can occur independently of each other. This suggests genome-wide remethylation just before or during that tumor-associated DNA hypomethylation contri- implantation throughout the early embryoblast (Chap- butes to carcinogenesis or tumor progression separately man et al., 1984; Oswald et al., 2000; Razin and Kafri, from aberrant DNA hypermethylation and its atten- 1994; Santos et al., 2002). Although there are dant silencing of tumor suppressor genes. The role of methylation changes later in embryogenesis, they are DNA hypomethylation in cancer may involve increas- much less wide-spread (Razin and Kafri, 1994). ing karyotypic instability (Mitelman, http:// Nonetheless, as for many types of cancer, there is the cgap.nci.nih.gov/Chromosomes/Mitelman, accessed possibility that during oncogenesis there was selection 2001; Qu et al., 1999b; Tuck-Muller et al., 2000; Wong for rare cells in the untransformed precursor tissue that et al., 2001) and also altering gene expression (reviewed had pre-existing hypomethylation or hypermethylation in Ehrlich, 2002). in the studied DNA sequences. The quantitative nature of the MethyLight assay for Materials and methods local methylation changes and the inclusion of diverse postnatal somatic tissues of normal origin make our Tissue samples and DNA isolation results on regional DNA methylation especially Wilms tumor samples (Table 3) were obtained during informative. For example, the hypermethylation of nephrectomy of untreated patients except for WT14A, which TNFRSF12 (DR3 or TRAMP) in more than half of the was a biopsy sample. Informed consent was obtained from cancers (Table 3) could not have been detected without the patients or unlinked samples were used, with IRB this quantitation, given the considerable levels of approval in all instances. This set of tumors is different from methylation of this sequence in normal tissues (Table the Wilms tumors described in our previous study (Qu et al., 2). TNFRSF12 is a member of the TNFR superfamily 1999b). The control tissues were autopsy samples from implicated in the apoptotic response to cellular insults trauma victims. DNA was purified from quick-frozen samples by standard methods (Ehrlich et al., 1982). (Wang et al., 2001). Although cancer-associated changes in methylation or mutagenesis of TNFRSF12 5 have not been reported, evidence for cancer-linked Quantitation of global m C levels in the genome and DNMT deletion of TNFRSF12 on 1p36 has been found RNA analysis (Grenet et al., 1998). The overall DNA m5C content was determined by high By the use of a quantitative DNA hypermethylation performance liquid chromatography (HPLC) (Tuck-Muller et assay, we avoided potential false positive results from al., 2000) on heat-denatured DNA digested with nuclease P1 over-amplification of extremely low amounts of and then bacterial alkaline phosphatase for 2 h at 378C each. methylated DNA sequences. This might explain the To quantitate the relative amounts of DNMT1, DNMT3A, discrepancy between our finding of negligible levels of and DNMT3B RNA in the different tumor samples, real-time RT – PCR was used on total RNA isolated and analysed as methylation of the 5’ CpG island of CDKN2A and the previously (Eads et al., 1999) with the exception that the report by Arcellana-Panlilio et al. (2000) of methyla- following primers and probes were used for the H4F2 RT – tion of this region in seven out of seven Wilms tumors PCR analyses: forward primer, 5’-TTCGGGACGCAGT- exhibiting decreased CDKN2A expression. They used a CACCTA-3’; reverse primer, 5’-AGCGCGTACACCACAT- non-quantitative methylation-specific PCR involving 45 CCAT-3’; and probe, 6FAM5’-CACGCCAAGCGCAAGA- cycles of amplification. In our quantitative assay, we CCGTC-3’ TAMRA. found extremely low amounts of methylation in this region (PMR 0.003 – 0.04; data not shown) for eight Quantitative analysis of methylation levels in CpG-rich regions out of nine assayed tumor DNAs. Given the of the genome exponential nature of PCR amplification, these tumors The bisulfite modification-based MethyLight assay was used might have appeared as positives in their assay. to quantitate methylation at CpG-rich regions (Table 4) Importantly, we found no association between the (Eads et al., 2000, 2001). The extent of methylation was extent of methylation of any of the 13 CpG-rich determined by amplification of bisulfite-treated genomic regions in the 31 Wilms tumors analysed and their DNA with primers and a probe specific for the bisulfite- global DNA methylation levels. A negative association converted, fully methylated sequence for each gene (Table 4). between CpG island methylation and global DNA To control for the amount of input DNA, this value was hypomethylation would have been predicted if tumors normalized to the extent of amplification of an ACTB DNA with much CpG island hypermethylation were less sequence lacking CpG dinucleotides (Eads et al., 2001). The likely to also display DNA hypomethylation. A percentage of fully methylated molecules at a specific CpG- positive association might have been seen if CpG rich gene region was calculated by dividing the MethyLight signal for the given gene in the sample by that for ACTB in island hypermethylation preceded and provoked global the sample DNA and then dividing that ratio by the genomic hypomethylation or vice versa. CpG island analogous GENE:ACTB ratio for in vitro methylated sperm hypermethylation and global hypomethylation have DNA and multiplying by 100. The value obtained is been reported to co-exist in cancers (Narayan et al., designated as the percentage of methylated reference (PMR) 1998; Ribieras et al., 1994; Ushijima et al., 1997), but (Eads et al., 2001).

Oncogene Wilms tumor DNA hypomethylation and hypermethylation M Ehrlich et al 6701 DNA hypomethylation, and for possible associations between Statistical analysis global hypomethylation and DNMT1, DNMT3A, and Samples of normal tissues (Table 1) were used for 10 DNMT3B RNA levels (Hollander and Wolfe, 1973). independent control observations from the oldest and youngest autopsy samples available. The non-parametric Wilcoxon rank sum test was used to examine possible diﬀerences in hypermethylation of tumor and normal DNAs and to test whether global DNA hypomethylation is Acknowledgments associated with regional hypermethylation in tumors (Hollan- We are very grateful to Michael Jaynes, Darla Tate, Shirley der and Wolfe, 1973). To test for a possible association Waldon, Martine Therrien, Michelle Maguigad, and Cheryl between satellite DNA hypomethylation and regional Medeiros-Nancarrow for invaluable help in collecting hypermethylation, the Kruskal Wallis test was used with tumor samples and obtaining clinical information. We the satellite DNA hypomethylation scored as 7,+, or thank Kazuko Arakawa for her assistance in statistical ++(Ehrlich et al., 2002). The test of Spearman correlation computing. This work was supported in part by NIH coeﬃcient=0 was used to look for possible associations grants CA 81506 (to M Ehrlich), CA 75090 (to PW Laird), between tumor stage and individual PMR values or global and PO1 CA 46589 and PO1 CA 70972 (to E Fiala).

References

Ahuja N, Li Q, Mohan AL, Baylin SB and Issa JP. (1998). Gama-Sosa MA, Slagel VA, Trewyn RW, Oxenhandler R, Cancer Res., 58, 5489 – 5494. KuoKC,GehrkeCWandEhrlichM.(1983).Nucleic Acids Arcellana-Panlilio MY, Egeler RM, Ujack E, Pinto A, Res., 11, 6883 – 6894. Demetrick DJ, Robbins SM and Coppes MJ. (2000). Genes Gardiner-Garden M and Frommer M. (1987). J. Mol. Biol., Chromosomes Cancer, 29, 63 – 69. 196, 261 – 282. Baylin SB and Herman JG. (2000). DNA Alterations in Gaughan DJ, Barbaux S, Kluijtmans LA and Whitehead AS. Cancer: Genetic and Epigenetic Alterations. Ehrlich M (ed). (2000). Gene, 257, 279 – 289. Natick: Eaton Publishing, pp. 293 – 309. GrenetJ,ValentineV,KitsonJ,LiH,FarrowSNandKidd Burbee DG, Forgacs E, Zochbauer-Muller S, Shivakumar L, VJ. (1998). Genomics, 49, 385 – 393. Fong K, Gao B, Randle D, Kondo M, Virmani A, Bader S, Harada K, Toyooka S, Maitra A, Maruyama R, Toyooka Sekido Y, Latif F, Milchgrub S, Toyooka S, Gazdar AF, KO, Timmons CF, Tomlinson GE, Maslrangel D, Hay RJ, Lerman MI, Zabarovsky E, White M and Minna JD. Minna JD and Gazdar AF. (2002). Oncogene, 21, 4345 – (2001). J. Natl. Cancer Inst., 93, 691 – 699. 4349. Busson P, Zhang Q, Guillon JM, Gregory CD, Young LS, Hollander M and Wolfe DA. (1973). Nonparametric Clausse B, Lipinski M, Rickinson AB and Tursz T. (1992). Statistical Methods. New York: John Wiley and Sons Int. J. Cancer, 50, 863 – 867. pp 68 – 75 and 191 – 194. Chapman V, Forrester L, Sanford J, Hastie N and Rossant J. Issa JP. (2000). DNA Alterations in Cancer: Genetic and (1984). Nature, 307, 284 – 286. Epigenetic Alterations. Ehrlich M (ed). Natick: Eaton ChengP,SchmutteC,CoferKF,FelixJC,YuMCand Publishing, pp. 311 – 322. Dubeau L. (1997). Br. J. Cancer, 75, 396 – 402. Issa JP, Vertino PM, Boehm CD, Newsham IF and Baylin Dreijerink K, Braga E, Kuzmin I, Geil L, Duh FM, Angeloni SB. (1996). Proc. Natl. Acad. Sci. USA, 93, 11757 – 11762. D, Zbar B, Lerman MI, Stanbridge EJ, Minna JD, Malik K, Salpekar A, Hancock A, Moorwood K, Jackson S, Protopopov A, Li J, Kashuba V, Klein G and Zabarovsky Charles A and Brown KW. (2000). Cancer Res., 60, 2356 – ER. (2001). Proc. Natl. Acad. Sci. USA, 98, 7504 – 7509. 2360. Eads CA, Danenberg KD, Kawakami K, Saltz LB, Mares J, Kriz V, Weinhausel A, Vodickova S, Kodet R, Haas Danenberg PV and Laird PW. (1999). Cancer Res., 59, OA, Sedlacek Z and Goetz P. (2001). Cancer Lett., 166, 2302 – 2306. 165 – 171. Eads CA, Danenberg KD, Kawakami K, Saltz LB, Blake C, Narayan A, Ji W, Zhang XY, Marrogi A, Graff JR, Baylin Shibata D, Danenberg PV and Laird PW. (2000). Nucleic SB and Ehrlich M. (1998). Int. J. Cancer, 77, 833 – 838. Acids Res., 28, E32. Oswald J, Engemann S, Lane N, Mayer W, Olek A, Fundele Eads CA, Lord RV, Wickramasinghe K, Long TI, R, Dean W, Reik W and Walter J. (2000). Curr. Biol., 10, Kurumboor SK, Bernstein L, Peters JH, DeMeester SR, 475 – 478. DeMeester TR, Skinner KA and Laird PW. (2001). Cancer Qu G, Dubeau L, Narayan A, Yu M and Ehrlich M. (1999a). Res., 61, 3410 – 3418. Mut. Res., 423, 91 – 101. Ehrlich M. (2000). DNA Alterations in Cancer: Genetic and Qu G, Grundy PE, Narayan A and Ehrlich M. (1999b). Epigenetic Changes. Ehrlich M (ed). Natick: Eaton Cancer Genet. Cytogenet., 109, 34 – 39. Publishing, pp. 273 – 291. Razin A and Kafri T. (1994). Prog. Nucleic Acid Res. Mol. Ehrlich M. (2002). Oncogene, 21, 5400 – 5413. Biol., 48, 53 – 81. Ehrlich M, Hopkins N, Jiang G, Dome J, Yu M, Woods C, Ribieras S, Song-Wang X-G, Martin V, Lointier P, Frappart Tomlinson G, Chintagumpala M, Champagne M, Diller L, L and Dante R. (1994). J. Cell Biochem., 56, 86 – 96. Parham D and Sawyer J. (2002). Cancer Genet. Cytogenet. Santos F, Hendrich B, Reik W and Dean W. (2002). Dev. in press. Biol., 241, 172 – 182. Ehrlich M, Gama-Sosa M, Huang L-H, Midgett RM, Kuo Shridhar V, Bible KC, Staub J, Avula R, Lee YK, Kalli K, KC, McCune RA and Gehrke C. (1982). Nucleic Acids Huang H, Hartmann LC, Kaufmann SH and Smith DI. Res., 10, 2709 – 2721. (2001). Cancer Res., 61, 4258 – 4265. Feinberg AP and Vogelstein B. (1983). Nature, 301, 89 – 92.

Oncogene Wilms tumor DNA hypomethylation and hypermethylation M Ehrlich et al 6702 Strichman-Almashanu LZ, Lee RS, Onyango PO, Perlman Wang EC, Kitson J, Thern A, Williamson J, Farrow SN and E, Flam F, Frieman MB and Feinberg AP. (2002). Genome Owen MJ. (2001). Immunogenetics, 53, 59 – 63. Res., 12, 543 – 554. Wong N, Lam WC, Lai PB, Pang E, Lau WY and Johnson Toyota M, Shen L, Ohe-Toyota M, Hamilton SR, Sinicrope PJ. (2001). Am.J.Pathol.,159, 465 – 471. FA and Issa JP. (2000). Cancer Res., 60, 4044 – 4048. Yung R, Ray D, Eisenbraun JK, Deng C, Attwood J, Tuck-Muller CM, Narayan A, Tsien F, Smeets D, Sawyer J, Eisenbraun MD, Johnson K, Miller RA, Hanash S and Fiala ES, Sohn O and Ehrlich M. (2000). Cytogen. Cell Richardson B. (2001). J. Gerontol. A Biol. Sci. Med. Sci., Genet., 89, 121 – 128. 56, B268 – B276. Tycko B. (2000). DNA and Alterations in Cancer: Genetic and Zhang X-Y, Loﬂin PT, Gehrke CW, Andrews PA and Epigenetic Alterations. Ehrlich M (ed). Natick: Eaton Ehrlich M. (1987). Nucleic Acids Res., 15, 9429 – 9449. Publishing, pp. 333 – 349. Ushijima T, Morimura K, Hosoya Y, Okonogi H, Tatematsu M, Sugimura T and Nagao M. (1997). Proc. Natl. Acad. Sci. USA, 94, 2284 – 2289.

Oncogene